1. Field of the Invention
This invention relates generally to optimization of optical communication networks and more particularly to the deployment of forward error correction (FEC) for optimization of parameters in the operation of photonic integrated circuits (PICs) in optical transmitters as depicted, for example, in FIG. 70 of the above identified parent application.
2. Description of the Related Art
It is an expectation for optical communication systems in optical transport networks to operate at a low bit error rate (BER), such as, 10−12 and lower. In the past, when the demands on such systems was slower due, for example, to larger optical channel spacing, negligible electro-optical element crosstalk and negligible network nonlinearities, all that was generally required was to increase the power on the transmitter channels in order to have negligible error rate at the optical receiver.
Thus, forward error correction (FEC) was not necessary in those times. However, as the transmission capacity of such systems has increased as well as the baud rate of signal transmission, together with denser wavelength division multiplexing (DWDM) and closer channel spacing and therefore higher crosstalk affinity, the deployment of FEC has become widely prevalent in optical transport systems today. The reason why that higher channel transmitted power cannot be employed is that optical fiber nonlinearities prevent increases in power. For example, four wave mixing (FWM) and amplified spontaneous emission (ASE) may become significant and deteriorate the channel signals. Second, employing FEC provides a cost-performance tradeoff. The use of FEC permits longer signal reach over the optical communication link before signal channel regeneration is a necessity. This is because the optical link can operate at lower received signal power to achieve the same BER specified or necessary for successful system operation. However, the tradeoff requires additional signal processing at the optical receiver to decode the FEC and provide signal correction for erroneous signal bits to achieve desired or lower BER. Thus, FEC is typically used to provide significant savings in the overall optical power budget of a fiber link, by allowing the system to operate at a much higher line BER. For example, the (255,239) Reed-Solomon code (cited above) corrects a 10−4 line BER to 5×10−15. With FEC, the optical link can support line BER rates up to about 10−4, report on the high BER in real-time, and correct the errors to better than 10−12, which is a typical target maximum BER for optical communication networks.
Forward error correction or FEC is a technique for using error-correcting code to reduce the bit error rate (BER) on a communication channel. The process involves the transmission of additional bits with the client signal to provide signal redundancy of the client signal data bits. These redundant bits are employed at the optical receiver to correct most of the errors found in the actual client signal data bits. This process therefore can enhance the BER at the optical receiver so that an acceptable or lower BER can be realized. Thus, in simple terms, FEC is a process for reducing the transmitted signal error rate by employing the transmitter to send (i.e., “forward”) redundant client signal bits using error-correcting code.
A recent advent in optical transmission equipment has been the advent of multiple channel optical transmitter photonic integrated circuit (TxPIC) and multiple channel optical receiver photonic integrated circuit (RxPIC) chips as disclosed and taught in U.S. patent application Ser. No. 10/267,331, filed Oct. 8, 2003, which application is owned by the common assignee herein and is incorporated herein by its reference, and U.S. Pat. No. 7,116,851, issued Oct. 3, 2006, supra, from which this application is a continuation-in-part. The TxPIC and RxPIC are monolithic chips, having multiple active and/or passive elements integrated on a single substrate. Reference to a PIC chip also includes integrated circuits known as a planar lightwave circuit (PLC). The TxPIC chip comprises an integrated array of modulated sources, defined as an array of directly modulated laser sources or an integrated array of laser sources integrated with electro-optic modulators. Each of the modulated sources represents a signal channel, each of a different emission wavelength, for generating an optical signal for transport in an optical communication network. The modulated sources have their outputs coupled to inputs of an integrated optical combiner. For example, the laser array may be DFB lasers or DBR lasers. The electro-optical modulator may be comprised of electro-absorption (EA) modulators (EAMs) or Mach-Zehnder modulators (MZMs). The optical combiner is preferably a wavelength selective combiner or multiplexer, where examples of such a wavelength selective combiner are an Echelle grating or an arrayed waveguide grating (AWG). The disclosure of this application illustrates many different embodiments of the TxPIC and RxPIC, applications of the TxPIC and RxPIC in an optical transport network and in wavelength stabilization or monitoring of the TxPIC and RxPIC. Such a monolithic TxPIC or RxPIC chip with integrated multiple signal channels and integrated wavelength selective combiner is the first of its kind anywhere in use in optical transport equipment as well as disclosed in the art.
The deployment of multiple channels in a PIC results in compromised performance vis-à-vis each other. Said another way, integrating multiple active and/or passive elements on a single chip requires performance tradeoffs between the elements. For example, one trade-off might be degradation in laser performance across the array of laser sources, such as in the case of threshold current, spectral characteristics, and operational efficiency as well as increase in relative intensity noise (RIN). Second, performance degradation across the transmission wavelength window may occur in the array of modulators. It should be understood that placing a practical number of operable wavelengths on a TxPIC to achieve worthwhile integration results in a range of laser wavelengths within a modulator operating window that is inconsistent with techniques and approaches employed in the past and is a monumental task. Besides transmitter impairments, a communication system must also deal with impairments on the receiver, or RxPIC, that may include, but are not limited to, semiconductor optical amplifier (SOA) and/or photodetector (PD) noise, noise figure (NF), polarization dispersion loss (PDL), polarization-dependant gain (PDG) penalties, etc.
Specific devices have been developed to help reduce some sources of signal noise. For example, an optical isolators, or Faraday isolators, is designed to allow transmission of a signal in only one direction, e.g., downstream from the transmitter, but block transmission, or reflections in the opposite direction, that would otherwise degrade the performance of upstream optical devices, such as the laser or modulator. However, if an optical isolator is located immediately after a laser or modulator to protect those optical devices from noise and reflection then integrating then integrating them into a monolithic device would be difficult as materials used in an optical isolator include ferro magnetic magneto-optical material. Consequently, a need arises to eliminate the need for an optical isolator, or to provide a monolithic device that does not require an optical isolator on the integrated circuit, e.g., that does not require an optical isolator be placed immediately after the laser or modulator.
In the fabrication of a dense WDM system on a single PIC chip, very precise wavelength control across the laser array integrated on the chip is important. In present conventional systems employing discrete laser sources or EMLs, this is not an issue because individual laser or EMLs can be binned and later mixed and matched at the system level according to their peak lasing wavelength. However, in the case of integrated arrays of lasers on a single chip, there is no luxury to mix and match because all of the optical signal sources are formed in large scale photonic integration in close relationship and, therefore, the performance of all the lasers in the array must be initially and successfully fabricated to desired specifications so that all of the integrated electro-optic elements on a PIC respectively meet the requisite assigned grid wavelength emission and operational specifications without unacceptable signal distortion deterioration if desirable yields of such photonic integrated circuits or PICs are to be realized. Thus, what is required is to initially achieve a sufficient wafer yield of sufficiently distortion-free transmission channels as well as wavelength control across an integrated circuit with a modulated source array comprising a plurality of modulated sources in order to achieve a dense WDM system on a semiconductor chip, such as an InP-based semiconductor PIC. In addition, after the demanding wavelength requirements are met, the other transmission properties (e.g., power, BER, optical channel noise, etc.) must be sufficient across all PIC signal channels for the intended optical signal transport application. While the TxPIC carries many different active and passive integrated components, by far, the laser sources, such as a DFB laser array, have the tightest wavelength specification requirements, compared with the EA modulator array, which may have a wider wavelength operation window. Said another way, the TxPIC yield will be a strong function of DFB yield, especially wavelength yield, where the array of DFB lasers for each PIC die are substantially operating at desired on-chip emission wavelengths and sufficiently free of signal distortion or deterioration. Similarly, a sufficient wafer yield of sufficiently distortion-free receiver channels is needed for integration of a WDM system on a semiconductor chip. In the RxPIC, receiver channel noise and cross-talk can also affect a received signal quality up to the point it is translated into the electrical domain for digital signal processing.
Thus, as indicated previously, having the optical channel signals generated on a monolithic, large scale integration, multi-channel PIC chip have some attendant issues relative to providing acceptable BER levels at the optical receiver. In particular, issues can arise from either the serial coupling of active/passive elements in a given channel (intrachannel) or the interaction between the active/passive elements from one channel to another (interchannel). Because the signal channels are physically much closer to one another on a PIC having multiple channels, such as compared, for example, to the use of discrete modulated sources and discrete wavelength selective combiner, and because they are generally interconnected by low-loss waveguides and passive elements, there is an increase potential of signal distortion or deterioration due to, for example, channel crosstalk, RIN, SMSR, optical feedback channel noise, occasional wavelength hops. Also, in connection with feedback channel noise, there is reflected feedback, such as optical feedback reflection from the wavelength selective combiner in the channel sources, which can also deteriorate channel signal quality.
Referring now to FIGS. 1A through 1D, graphs of channel performance for a multichannel communication device/system and the resultant BER channel performance are shown. FIG. 1A graph 100 illustrates a hypothetical performance versus wavelength profile 102 for a monolithic integrated circuit having multiple channels for traversing wavelengths λ1 through λN where ‘performance’ can be any performance metric, e.g. gain, dispersion performance, etc. FIG. 1A shows a performance minimum on λ1, while the balance of the channels has approximately equal performance. FIG. 1B provides an exemplary BER by channel graph 120, corresponding to the channel performance profile shown in FIG. 1A. That is, the minimum performance in channel for λ1 of FIG. 1A may appear as an unacceptable BER, per threshold 104, on the same channel, λ1, shown in FIG. 1B, or on a different channel such as might arise from cross-talk. While FIGS. 1A and 1B illustrate a performance/BER channel coupling for a monolithic IC communication device, FIGS. 1C and 1D illustrate a similar performance/BER channel coupling for an optical line amplifier. FIG. 1C graph 140 illustrates an EDFA gain (G) vs. wavelengths λ1 through λN. Unfiltered gain profile versus wavelengths 142 can be compensated by a flattening filter, as is know to those skilled in the art, thereby producing a resultant flattened gain profile 144 that is typically not completely flat and linear, as shown. Similar to FIG. 1A, FIG. 1C shows different channels performing differently on gain, thereby resulting in a max performance on channel λ2, and a minimum performance for channel for λN-1. FIG. 1D provides the exemplary BER v. wavelength graph 160 that corresponds to the performance versus wavelength shown in FIG. 1C in channel. Thus, the minimum gain channel, λN-1, corresponds in this example to a maximum BER that exceeds BER threshold 146. Because of the failure of λN-1 to meet the threshold BER, the communication device responsible for the poor performance, e.g., a TxPIC and/or an RxPIC, may be unsuitable for field application, despite the satisfactory performance of the balance of the channels on the chip. Consequently, a need arises for a means to recover a chip with one or more channels that fail to meet BER, or other performance metrics, when the balance of the channels are satisfactory.
Referring now to FIG. 2, a schematic view of a communication network 200 with multiple sources of noise is shown. Many sources of noise and signal distortion arise in a communication network. Sources can include, but are not limited to noisy PICs 205-208, acting as either transmitters or receivers or transceivers (e.g. PIC 205 is TxPIC to PIC 206 as RxPIC, while PIC 207 is RxPIC to Pic 208 TxPIC, thus providing), that are coupled via a communication channel, e.g., bidirectional fiber link 202A and 202B, themselves causing noise and distortion and from environmental factors on the channel 203. A cumulative summation of noises from all sources, Σi=0n noise(i) determines whether the received signal together the noise and distortion Input signal on line 204 is a clean electrical signal from a local client, from which PIC 205, as a transmitter, generates an optical signal for transmission on the line fiber link 202, e.g. an optical fiber. Channel noise and interference may arise from optical couplings, splices, taps, in-line amplifiers, compromised cladding, and other environmental influences. An ideal system would generate an output signal from transmitter 205 with extremely low noise or distortion. However if a noisy transmitter generates an unusually high amount of errors, it may consume the ability of a per-channel FEC to recover the data. Consequently, a need arises for a system to compensate and recover data when a transmitter or receiver are responsible for high error rates or signal distortion.
Referring to FIGS. 3A through 3C, graphs illustrating sources for signal degradation based on channel parameters are shown. FIG. 6A graph 300 illustrates compact channel spacing 302 grid. Because of the proximity of adjacent channels, e.g., λ1-λ2, potential interference 304 on the fringes of the channels might result in destructive signal interference, and hence bit errors, during communication. FIG. 6B illustrates channel performance graph 320 resulting from optical impairments, that potentially could lead to phase shift, frequency offset and drift, that could produce actual interference 326, despite an apparently sufficient guard band based on the nominal channel spacing 324. FIG. 6C illustrates a graph 340 of bandwidth goals for a photonic chip. The effective bandwidth of a convention chip is constrained to nominal bandwidth 342, due to non-linear or non-scalable performance of active and/or passive devices, as well as to process variations and optical impairments during fabrication. Future bandwidth requirements may be extended to accommodate growth in data transport rates. Consequently a need arises to overcome the errors generated by close channel spacing, manufacturing, or fabrication optical impairments, and bandwidth nonlinearities in order to meet future demands for expanded operating envelope.